Burst Detection Checklist for GPS 874465554 event
The LSC-VIRGO Burst Working Group
October 1, 2007

The most recent version of this page in CVS

Top level event information available at http://lancelot.mit.edu/~lindy/s5/events/874465554.6807.html

Please notice that some of the suggested checks may conflict
with the notion of keeping ourselves blind from the possible
blind injection. Do not undertake any such checks until after
the blind injection challenge has been declared over.

Status: X (Complete), I (Incomplete), O (Open, Critical)

ID Category Check description Status Contributors Link Date Signed
1
Zero-level sanity
Convert GPS times to calendar times and check for suspiciousness.
Sep 22 2007 03:05:40 UTC not suspicious
X
Lindy
summary
Oct 3, 2007
2
Zero-level sanity
Read e-logs for the times/days in question.
Was there anything anomalous going on?
seismic activity and loud/long transient at H1; nothing at L1
X
Lindy
LHO
LLO
Oct 4, 2007
3
Zero-level sanity
Check state vector of all instruments around the time of the candidate.
Data correctly flagged?
FOM shows all IFO's in science mode
X
Nick
plots
Oct 10, 2007
4
Zero-level sanity
Identify inspiral range of the instruments in order to set
the scale of sensitivity. Is this typical/low/high?
Good H1/L1 range, typical H2 range. Range drop 6 min after
X
Nick
plots
Oct 10, 2007
5
Zero-level sanity
Identify nearest in time hardware injections (also
type/amplitude of them). Were there ongoing stochastic/pulsar
ones? When did they start and what was their amplitude?
Closest hardware burst injection 6 min after event, pulsar on
X
Lindy
log
Oct 3, 2007
6
Zero-level sanity
How close to segment start/end for all instruments does this event occur?
Science segment boundary is over 10,000s away for all instruments
X
Lindy
summary
Oct 3, 2007
7
Data integrity
Check for undocumented/unauthorized/spontaneous hardware injections.
Nothing in injection log. Nothing unexpected in any excitation channel
(but did not look at LSC-ETMY_EXC_DAQ, of course!)
X
Lindy
Peter Sh
summary
summary
Oct 3, 2007
Oct 9, 2007
8
Data integrity
Examine all possible test points recorded in frames or saved on disk.
The latter part might be time critical if data are overwritten.
9
Data integrity
Establish if there was any data tampering.
O
10
Data integrity
Check integrity of frames; check raw/RDS/DMT frames.
data-valid flags are OK on the L1 RDSs for both observatories
no CRC mismatch over range [874463712, 874467392)
X
John Z
Ben J
CRC check
Oct 9, 2007
11
The event
Run Q-scan on RDSs/full frames and on all available instruments
in the network of detectors.
Some mildly interesting things (see link).
H1 is very glitchy accross IFO channels, must understand
H2 is at inconsistent Q of 23, while 4km IFO's have Q of 5
X
Lindy
Leo
Romain
GW/RDS
discussion
full qscan
analysis
Qevents
Oct 3, 2007
12
The event
Run Event Display on RDSs/full frames and on all available instruments
in the network of detectors.
no obvious glitches. some spikes in AS-AC,REFL-I/Q about 1 secs later.
X
Shantanu
Event-Display
Oct. 5 2007
13
The event
What is the overall time-frequency volume of the event in each detector?
Does it look consistent? minimal uncertainty, looks very consistent
X
Lindy
GW qscan
Oct 3, 2007
14
The event
What is the expected background for such a candidate?
What is the significance of the observation given the background?
Compare background estimated from time-slides and from
first-principle Poisson estimates. Is it consistent?
background depends on algorithm/search. at the predetermined tresholds
for the online search we have a significance of about .03 for KW+CP.
However the event is well above the set thresholds. This is for
the probability of observing N events on Sept 21, 2007
O
Lindy
Erik
Igor
summary more
Oct 4, 2007
15
The event
How robust is the background estimate? Do randomly-chosen shifts
as opposed to fixed shifts- is the result consistent? How about other steps?
currently the background is only estimated for Sept 21- when different strides
and higher statistics (more slides) used, estimate remains consistent with statistics
X
Erik
Lindy
summary
Oct 12, 2007
16
The event
How robust is the significance of the event to the threshold
chosen? Do a "stair-case" analysis (varying threshold) to
appreciate this.
Naturally the significance of the event depends on the threshold
For KW+CP it varies from .03 (fg 3, bg .68) to .01 (1, .01)
depending on Gamma threshold chosen (5 or 9.5)
X
Lindy
summary
Oct 4, 2007
17
The event
If more than 2 IFOs are involved in the event, would any
2-IFO pair be able to identify it as well and with what
background/significance? easily visible in H1L1.
probably visible in H1H2 Q analysis (guessing from qscan)
X
Lindy
Oct 3, 2007
18
The event
If only 2-IFOs are involved, why it was not detected by the
rest of the detector network?
below KW threshold in H2, but clear signal in all 3 IFO's
G1 and V1 too poor in sensitivity at 100 Hz
X
Lindy
Oct 3, 2007
19
The event
Examine the status of the detectors not involved in the event
and establish why this is so.
H1H2L1G1V1 all in Science Mode (wow)
X
Lindy
Oct 3, 2007
20
The event
Examine frequency content out to the KHz range- is there anything
there? Will a broader bandwidth search pick it up?
There is nothing in h(t) up to 8192 Hz. However the high
frequency excess in AS_Q_0FSR for both H1 and L1 visible in
in the qscans should be understood (aliasing at nyquist?)
No triggers near the event time stand out in the high frequency search.
X
Lindy
Brennan
Leo
summary
HF triggers summary
HF qscan
Oct 4, 2007
21
The event
Run parameter estimation code.
Note: The parameter estimation code in E-D maybe outdated
X
Shantanu
Right-most column of E-D
Oct. 19, 2007
22
The event
Identify the sky ring/sky patch using non-coherent methods,
i.e., signal timing and amplitude
Calculated using time delay between H1 and L1
X
Adam
skymap (slide 4)
23
The event
Compare source reconstruction between coherent and incoherent methods
Sergei
Adam
24
The event
What fraction of the network's acceptance comes from this direction?
X
Sergei
Adam
skymaps
25
The event
How robust is the candidate against the stride size/start of the analysis?
Seen in KW and QPipeline, so it should be robust
X
Lindy
summary
Oct 4, 2007
26
The event
How the peak times established by the method, parameter estimation
and the various coherent methods compare? Is this as expected?
using QPipeline as a reference, KW peak is 9ms later. Both algorithms
give dt=4ms between H1 and L1. This is not at all bad for something
at the low frequency of 100 Hz.
CorrPower peak is at -38ms with a 50ms window
X
Lindy
summary
Oct 4, 2007
27
The event
Could it be an artifact of the prefiltering?
Q Pipeline and CorrPower use zero-phase filtering. The positive
offset of the KW trigger is due to an uncorrected filter delay
X
Lindy
Oct 4, 2007
28
The event
What is the detection efficiency for the search method
calculated near to the candidate event? average/low/high w/r/t the S5?
This is a job for all methods - we need to understand the range of
the network for the sky location at the the time of this event
O
Lindy
29
The event
Could there be any effect from lines not filtered enough and/or
any other artifact? Could it be violin mode excitations? Any other
mechanical resonances sneaking in?
does not look like a line in qscan. short duration (few ms)
X
Lindy
summary
Oct 4, 2007
30
The event
How stationary were the instruments around the time of the event?
Quantify this both in terms of singles counting and PSD.
No excess in low threshold KW rates. The spike in the high
threshold rates nearby (at the 11th hour) are hardware injections
I
Erik
Soma
Robert St
Lindy
summary
KW trend
NoiseFloorMon H0
NoiseFloorMon L0
Oct 8, 2007
31
Vetoes
What could have caused this event other than astrophysics?
blind injection. faulty injection. common low-freq glitch
O
Lindy
Oct 3, 2007
32
Vetoes
Are there obvious environmental disturbances in the Q-scan/Event Displays?
nothing obvious in the initial RDS qscans. however, we might want to
look at qscans which display all channels regardless of significanct
content immediately around event time. There is a nearby low freq
glitch in H0:PEM-BSC3_ACCX at -200ms, and noise in
H0:PEM-BSC6_MAGZ from -6 to 1s about event time
H0:PEM-BSC3_ACCX started experiencing excess noise at low freq
since Aug 26th. This is assumed to be non-physical and is
inconsistent with other accelerometers (see elog link)
X
Lindy
summary
BSC3_ACCX elog
Oct 4, 2007
33
Vetoes
Are there obvious interferometric disturbances in the Q-scan/Event Displays?
H1 shows a variety of IFO channels glitching nearby stopping 1s
after event time. The glitches are not directly in coincidence with
the event, but should be understood
X
Lindy
summary
Oct 4, 2007
34
Vetoes
Examine what known earthquakes occured around the time of the event.
fairly quiet around time of the event
X
Soma
Robert St
iris log
Oct 8, 2007
35
Vetoes
Contact power companies and obtain known power line transients
around the time of the event.
Robert
Oct 8, 2007
36
Vetoes
If available, check in/out records for trucks and heavy equipment
to the Lab that might not have been recorded in the ilog. Check
for overpassing airplanes (airport flight logs).
Robert
Oct 8, 2007
37
Vetoes
Check for any switching of major electrical equipment around the time
of the event that might not have been recorded in the ilog
Robert
Oct 8, 2007
38
Vetoes
What are the KleineWelle (KW) triggers in auxiliary channels
around the time of the candidate?
Nothing much immediately around event time (ms within .714/.718)
Glitches within the same second are weak. BSC3_ACCX is missing in H0
as KW triggers are only generated from 10-512 Hz.
X
Lindy
Erik
by channel
by time
by signif
parameters
Oct 9, 2007
39
Vetoes
If there are any overlaps with KW trigger from auxiliary channels,
what is the expected background of such coincidence and what is the
significance of that channel as a veto channel?
Only BSC3 ACCX which is not significant
X
Erik
Lindy
veto analysis
Oct 12, 2007
40
Vetoes
Which of the overlapping channels are safe, which are not?
Analyze most recent hardware injections.
Not relevant
X
Lindy
41
Vetoes
For PEM/AUX channels with measured transfered functions, is the
signal present in the them consistent with the one in the GW one?
no expected DARM_ERR signal
X
Robert
Oct 10, 2007
42
Vetoes
If nothing in the non-GW channels in the RDS, proceed with scanning
full frames and repeat above checks.
43
Vetoes
Any known data quality flags overlapping with event? How is this
dependent on DQ flag threholds? What is the coincidence significance?
None of the DQ flags that have been evaluated so far overlap,
but many flags have not yet been evaluated for this part of the run
I
Peter Sh
segwizard
light dips
overflows
Oct 9, 2007
Oct 10, 2007
Oct 10, 2007
44
Vetoes
Examine minute trends/Z-glitch/glitch-mon data.
No suspicious features in trends or online DMT glitches
X
Keith R
Alan W
DMT glitches
Trends
Oct 17, 2007
Oct 25, 2007
45
Coherent Analysis
Run the H1-H2 Q analysis; anything in the H1-H2 null stream?
H1-H2 stream has norm energy of 7. H2 has energy of 15.
So H1 and H2 seem more or less consistent.
A careful look at the Q event display shows that the small signal
in the H1-H2 null stream is a result of an additional part of the
H2 signal slightly before the event. This could in principle be
an H2 glitch that happens just at the same time and frequency as
a real event or blind injection, though that seems unlikely.
To check this, 100 simulations are done where the best match
SG is injected coherently in H1 and H2. The qevents of the
injections show that while it isn't too uncommon to get
glitches in the null stream for a simulated event, none come
quite as close to the real candidate event.
I
Lindy
Shourov
GW qscan
injections
Oct 10, 2007
46
Coherent Analysis
Run the r-statistic cross-correlation over all detector pairs
(involved or not in the trigger); how significant each is?
47
Coherent Analysis
Run coherent analysis/null stream burst analysis on the
available detector network.
RIDGE pipeline finds maximum statistic at location of event
X
Sergei
Adam
Soumya
Kazuhiro
RIDGE
LIGO CED
LIGO/Virgo CED
analysis
skymap
Oct 8, 2007
48
Coherent Analysis
Run the inspiral multi-detector coherent analysis on the available
detector network and compare to the burst one.
Inspiral analysis shows excellent coherence, coherent SNR=13.3
X
Sukanta
Inspiral
Oct 23, 2007
49
Coherent analysis
What is the best fit waveform extracted from the data?
X
Sergei
Adam
LIGO CED
LIGO/Virgo CED
50
Other methods
Do other burst ETGs find the event(s)? If yes, compare
extracted event parameters, including background/significance.
The event is seen in all 3 IFO's using QPipeline, however no
coincidence is done, nor thresholds set for a detection
Event also seen in Block-Normal in H1 and L1 only
Event seen in CWB at FAR 1/598 days over S5A.
X
Lindy
Shantanu
summary
Block-Normal
CWB analysis
Oct 5, 2007
51
Other methods
What is the outcome of the Inspiral and/or Ringdown search
around the time of the burst event? If something is present,
what is the background/significance?
No online BNS event (1-3 Msun);
Offline playground inspiral search finds H1L1 coinc, masses>17.5 Msun
X
Lindy
Duncan

Inspiral slides
Oct 23, 2007
52
Calibrations
What is the calibration constants and errors around this event?
O
53
Calibrations
How robust is the event analysis against calibration version?
KW uses uncalibrated data. May affect CorrPower.
I
Lindy
Oct 4, 2007
54
Calibrations
Could there be any calibration artifact?
such a strong signal is not present in the excitation channel
X
Lindy
summary
Oct 4, 2007
55
Calibrations
Is the event identified when analysis is run on ADC data?
Compare findings of an analysis starting with ADC(t) vs h(t).
event is found on DARM_ERR with KW and h(t) with QP/CP
X
Lindy
Oct 4, 2007
56
Miscellanea
Check timing system of the instruments (well in sync?).
Szabi
Oct 8, 2007
57
Miscellanea
Check for any recent reboots, software updates/reloads.
Any suspecious acquisition software changes?
58
Miscellanea
Check recent logins to the various acquisitions computers.
59
Other GW detectors
Any signature in the non LSC-VIRGO detectors? TAMA/bars online?
TAMA not online
X
Kazuhiro
Oct. 16, 2007
60
Other GW detectors
What is the expected signal size given what we know for the event?
undetectable in V1, G1 and anything else due to low freq (100Hz)
X
Lindy
Oct 4, 2007
61
non-GW detectors
Any known or "sub-prime" event in E/M or particle detectors around the globe?
Not so far. Nearest GRB around this event is at 874357486 and 874610137
O
Ray, Isabel
S5 GRB list
Oct. 19, 2007
62
Astrophysics
Any known sources overlapping the ring/patch on the sky
corresponding to the direction of the candidate event?
Yes. ~60 potential hosts within 20 Mpc, including M33.
Nearby there is M31, M33, and LMC. Between 50-80 Mpc there is the
Perseus-Pisces supercluster. Between 1 and 50 Mpc there is not so much
an excess of galaxies at first glance, but of course there are
galaxies overlapping the possible sky positions.
I
Jonah
Lindy
Source Reconstruction
Astrophysical sources
63
Astrophysics
Examine events (other than the candidate) reconstructed at the
same direction. Perform a directional search; if a point source
is behind this, more, lower SNR events might be in our data.
64
Astrophysics
How the extracted waveforms compare to astrophysical waveforms?
What is the energy scale going into GW, assuming galactic distances?
O
Soumya
Kazuhiro
Oct 8, 2007
65
Vetoes
Create a hardware injection starting with signal waveforms
corresponding to the best fit waveforms extracted from the instruments.
66
Other methods
Take the extracted waveform per IFO and run matched filtered
search in order to establish how often the specific morphologies
appear in the data.
O
Myungkee
Soumya
Kazuhiro
analysis
RIDGE
Oct 16, 2007
67
The event
Run the Q-event display
The Q-event display shows optimal sum and H1H2 null streams
H2 appears to have two nearby blobs of singal energy, the second
is coherently aligned with the signal in H1, and the first is
completely independent. This can be seen by looking at the
incoherent null stream where the signal energy in H1 and H2 is
simply summed together compared to the null stream where H1-H2
is computed first. The incoherent blob in H2 is not significant
enough to qualify as a null stream veto (threshold of 10)
X
Shourov
Lindy
Q-event display
analysis
Oct 12, 2007
68
The event
Run the Coherent-Event-Display (CED)
X
Sergei
Adam
LIGO CED
LIGO/Virgo CED
Oct 8, 2007
69
Vetoes
Play audio files corresponding to GW, H1+-H2, auxiliary channels
Event audible in all three ifos with suitable filtering and
playback equipment: H2 faint, L1 and H1 distinct "heartbeat".
H1 has second "heartbeat" ~ 0.6 sec after first.
X
Peter Sa
Josh
Jess
summary
Oct8,2007
70
Vetoes
Check that signal is the same in all photodiodes.
Not seen in PdNMon, so should be same in all photodiodes
X
Robert
John Z.
Looks ok
Oct 19, 2007
71
Vetoes
Check wind speeds
Wind speeds normal and less than 15 mph
X
Shantanu
No 12 above
Oct. 8, 2007
72
Vetoes
Check for fluctuations in power levels of TCS laser
Normal: no mode hops or big jumps in TCS power levels
X
Shantanu
Summary
Oct. 8, 2007
73
Vetoes
Do a seismic Q-scan
Some seismic noise in L0:EY_SEISY and H0:EY_SEISZ
but probably consistent with background.
X
Shantanu
Romain
LHO  LLO
Q-scan followup
Oct. 15, 2007
74
Vetoes
Check on laser safety passcard records to see if anyone
was in a site LVEA or outbuilding during or near the event
75
The event
Check the QOnline analysis around the event
Nice triggers for H1 and L1
X
Shourov
Lindy
G1 H1 H2 L1 V1
Oct 15, 2007
76
Vetoes
Check for cosmic ray events in LIGO's CR detectors
The nearest event in time was about 6 minutes before,
at 9/22 02:59:55.887 UTC. The nearest after the event was at 03:35 UTC.
X
Ray
Oct 15, 2007
77
The event
Study time dependence of the signal frequency; does it vary?
HHT analysis of the event shows that, in H1 and L1, the frequency
falls over time by an amount which is significant compared to the
uncertainties in the frequency determination.
X
Alex
H1, L1
(nothing clear seen in H2)
May 20, 2008
78
Vetoes
Verify that auxiliary channels are working properly.
Check rate and BLRMS minute trends for suspicious activity.
Lindy
John Z.
June 19, 2008
79
Vetoes
Check level of upconversion in the instruments.
the equinox event occurred during
a period of typical levels of upconversion.
X
Robert
link
Sept 29, 2008